Drp1 and the cytoskeleton: mechanistic nexus in mitochondrial division

Abstract

Dynamin-related protein 1 (Drp1), the master regulator of mitochondrial division (MD), interacts with the cytoskeletal elements, namely filamentous actin (F-actin), microtubules (MT), and septins that coincidentally converge at MD sites. However, the mechanistic contributions of these critical elements to, and their cooperativity in, MD remain poorly characterized. Emergent data indicate that the cytoskeleton plays combinatorial modulator, mediator, and effector roles in MD by ‘priming’ and ‘channeling’ Drp1 for mechanoenzymatic membrane remodeling. In this brief review, we will outline our current understanding of Drp1-cytoskeleton interactions, focusing on recent progress in the field and a plausible ‘diffusion barrier’ role for the cytoskeleton in MD.

A historical perspective

Drp1 belongs to the dynamin superfamily, a cohort of evolutionarily linked, self-assembling, mechanoenzymatic GTPases (collectively termed dynamin superfamily proteins or DSPs), with well-established functions in various membrane remodeling events, including organellar fission and fusion (1). Remarkably, dynamin, the family namesake and founding member (though evolutionary recent in origin), was first isolated as a MT-associated protein with nucleotide-dependent MT-bundling and -remodeling properties (2). Its implication in vesicle scission during endocytosis came afterwards with the observation that it also self-assembled into helical polymers around the cylindrical membrane necks of deeply invaginated endocytic pits (3).

Likewise, columnar arrays of F-actin were found to sheath and encapsulate dynamin-coated membrane invaginations prior to endocytic vesicle scission (4). In addition, dynamin was found to be necessary for the formation of actin comets that drive vesicle release and propulsion into the cytosol (5). Whether these dynamin-cytoskeleton interactions were direct, or were essential for endocytosis, remained controversial (6,7). In time, similar cytoskeletal involvement was also discovered for Drp1-catalyzed MD (8-13).

Thus, the notion that DSPs are primarily cytoskeleton binders and remodelers revealingly pre-date their consequent connection to mechanoenzymatic membrane remodeling. Regardless, the functional significance and utility of DSP-cytoskeleton interactions were largely overlooked, or were deemed as non-essential, in favor of the more readily accessible membrane remodeling properties of the DSPs during their respective biological schemes.

Spatiotemporal positioning of the cytoskeleton in MD

MD, which counterbalances mitochondrial fusion events within the dynamic mitochondrial network occurs at defined ‘hot spots’ on the mitochondrial surface marked by the juxtaposition of ER-derived membrane tubules that transversely crossover and seemingly circumnavigate the organelle (14). This ER-mitochondria contact is thought to impose a narrow, local, cylindrical geometry on the mitochondrial outer membrane (MOM) amenable for constriction and fission (14,15). Coincident with this apposition, mitochondrial inner membrane (MIM) fission is thought to occur independently of the MOM (16,17).

MT arrays and associated molecular motors (kinesins and dyneins) not only anchor the elongated and dividing mitochondrial segments (18,19), but also support, transport, and facilitate the mitochondrial crossover of ER tubules (20). Dynamic F-actin networks and associated myosin motors occupy the interstitial space between the ER tubule and mitochondria fastening themselves onto either membrane via specific adaptor proteins (INF2 at the ER and Spire1C at the MOM) (9,21,22). In addition, septin filaments also intersect ER-mitochondria contact sites (23) and appear to align themselves on either side of the centrally located F-actin bundles, thus bookending the constricted MD site (24). Furthermore, septin bundles interact directly with, and guide, MT to MD sites, thus coordinating both F-actin and MT dynamics (24,25). Interestingly, INF2, which nucleates F-actin assembly at the ER-mitochondria interface, also interacts with MT (26). Thus, these three cytoskeletal elements, alongside their accessory molecules, appear to be physically and functionally coupled at MD sites.

The current model of MD invokes the hierarchical involvement of actomyosin networks at the ER-mitochondria interface, and subsequent Drp1 recruitment to these sites, to generate the progressive, high local membrane curvature (from ~0.5-1 μm overall diameter to <4 nm local diameter) requisite for membrane fission (14,27-30). Helical Drp1 polymers presumably self-assemble around MD sites pre-constricted by F-actin networks (to <200 nm local diameter) and undergo a GTP hydrolysis-dependent, force-generating conformational change that drives further membrane constriction (from 200 nm to <4 nm diameter) toward complete, leak-free fission (31-33). However, the molecular mechanisms underlying F-actin contribution to initial membrane curvature generation remain poorly described. How F-actin and Drp1 collaborate with MT and septins to effect cooperative membrane remodeling toward MD still remains unexplored. In the following sections, we will break down Drp1 interactions with each of the above three cytoskeletal elements from various biochemical and biophysical perspectives. We will also propose an overarching model that ties down their cooperative roles and plausible mechanisms in MD.

Drp1-F-actin interactions

Pioneering work from the Higgs laboratory has provided critical insights into Drp1-F-actin interactions during MD. F-actin networks at the ER-mitochondria interface have been implicated in the oligomeric maturation and assembly-competence of Drp1 for ensuing MOM recruitment and membrane remodeling (34). F-actin has been shown to directly bind Drp1 (34), and Drp1 recruitment, in turn, has been demonstrated to induce robust F-actin bundling (35). F-actin apparently recruits higher-order Drp1 oligomeric forms (> tetramers) from the cytosol, which exist in dynamic equilibrium with lower-order species (i.e. tetramers and dimers) in bulk solution (34,35). In addition, the GTP-bound Drp1 state (favoring self-assembly) is preferentially stabilized over the GDP-bound state (favoring disassembly) (35). Recruitment by F-actin (and ensuing Drp1-mediated F-actin bundling) drives further Drp1 self-assembly and stimulates Drp1 cooperative GTPase activity, resulting in dynamic Drp1 assembly-disassembly cycles over bundled F-actin (35). Inhibition of F-actin assembly by latrunculin B, or F-actin disassembly by cytochalasin D disrupts Drp1 recruitment and MD (8,9,13). These data indicate that initial F-actin-Drp1 interactions at the ER-mitochondria interface precede, and likely permit, Drp1 MOM recruitment (synergistically via MOM-anchored receptors such as Mff (34)) for MD.

Nevertheless, how F-actin interacts with Drp1, and the mechanisms underlying Drp1 maturation for mitochondrial recruitment still remain unclear. The four-helical bundle ‘stalk’ of dynamin comprised of the middle (MD) and GTPase effector domain (GED) regions was previously implicated in direct F-actin binding (36,37) (Fig. 1A). Yet, recent cryo-ET data have demonstrated that the disordered proline-rich domain (PRD) at the dynamin C-terminus, also enriched in arginine residues, instead binds F-actin, and induces F-actin bundling at the outer rim of the helical dynamin polymer (38). This is consistent with previous observations of F-actin arrays sheathing dynamin-coated membrane invaginations in vivo (4). Remarkably, however, Drp1 lacks the PRD (1). Instead, in place of the structured pleckstrin homology domain (PHD) in dynamin, Drp1 contains an intrinsically disordered region of ~111-148 residues termed the variable domain (VD) (39) (Fig. 1A). Similarly enriched in positively charged residue (K/R) clusters (Fig. 1B), the Drp1 VD presumably is the corresponding site for electrostatic F-actin interactions, as for the dynamin PRD (38). However, this is yet to be determined.

Figure 1.

(a) Domain organization in Drp1 versus dynamin. Domain arrangements within their primary structures are shown on the left and color-coded 3D structures are shown on the right. The Drp1 VD is intrinsically disordered and likely occupies the same position as the structured PHD in dynamin. (b) Sequence of the Drp1 VD highlighting positively charged residues in blue (+) and denoting the location of alternatively spliced segments. The ‘RRET’ sequence found in select splice variants and implicated in phospho-regulated direct MT interactions is underlined. Clusters of positively charged residues are identified numerically. These are potential sites of electrostatic interactions with F-actin, MT and target membrane lipids (e.g., CL and PA), whose surfaces are all negatively charged. (c) Cartoon illustration of our proposed ‘diffusion-barrier’ model of cytoskeleton function in MD. Drp1, conformationally primed by direct MT interactions at ER-mitochondria contact sites (MD sites), is sequestered together with its receptor (Mff) and target lipids by bounding F-actin bundles and septin filaments that either directly (septins) or indirectly (via ABPs) bind membranes, building ‘dams’ that contain the MD reaction components. Inset shows coexistence of positive and negative high membrane curvature regions at locally constricted MD sites.

Drp1-MT interactions

Conversely, the Drp1 VD is a known site of MT interactions. The VD is subject to post-transcriptional alternative splicing that gives rise to multiple Drp1 variants (isoforms) and to various post-translational modifications (PTMs) that differentially regulate Drp1 localization and function. Seminal work from the Strack laboratory identified a eleven residue-long segment within the VD, found in select splice variants and containing a pair of conserved R residues as part of a putative phosphorylation motif (RRE(T/S)), as the primary determinant of electrostatic Drp1-MT interactions (Fig. 1B) (10). Consistently, a phosphomimetic RRET to RREE mutation displaced Drp1 from MT and increased Drp1 mitochondrial localization and fission (10). Nevertheless, even the minimal VD segment (~111 residues; found in the shortest splice variant) contains numerous clusters of positively charged residues along its length that could potentially interact with the negatively charged MT surface (Fig. 1B). It is therefore unclear why stable Drp1-MT interactions are restricted to select Drp1 isoforms. It is likely that the various Drp1 splice variants all bind MT but have differential affinities (and turnover) for MT association, the biochemical and biophysical aspects of which remain to be determined. Moreover, how Drp1 is organized on MT, and how Drp1 activity in turn induces MT remodeling (MT bundling/restructuring), as previously described for both Drp1 and dynamin (10,40) remain a mystery. Thus, Drp1-MT interactions await elucidation at both structural and functional levels.

Drp1-septin interactions

In comparison, much less is known about Drp1-septin interactions. Septins, like Drp1, are self-assembling GTPases, but instead form ordered, curvilinear (non-helical) filaments (41). They are considered to be the intermediate filament-like fourth component of the cytoskeleton, due in part to their apolar polymerization and relatively poor enzymatic activity (41). Initially discovered at the septum during yeast cell division (hence the name), they are nevertheless widespread through the eukaryotic kingdom participating in multiple different intracellular processes, including the regulation of F-actin and MT dynamics. In the context of MD, septins interact with Drp1 in an F-actin-dependent manner (23), and have been shown to preferentially localize to regions of membrane curvature and bind negatively charged lipids via a polybasic region (PBR; α0 helix) adjacent to their G domains (41). Thus, septins likely function to couple F-actin, MT, and Drp1 dynamics to constricted MD sites that feature both positive and negative high membrane curvature (Fig. 1C). Septin depletion adversely affects Drp1 mitochondria recruitment and MD (23). Regardless, the molecular basis of Drp1-septins interactions and their essential functions in MD remain unexplored.

Plausible roles and mechanisms of Drp1-cytoskeleton interactions in MD

How does the cytoskeleton cooperate with Drp1 at MD sites to catalyze membrane fission? Current discussions of cytoskeleton involvement in MD are limited to the provision of membrane tension via cytoskeletal anchoring, and the creation of high local membrane curvature (membrane constriction) upon associated motor activity (i.e. myosin-dependent contraction of F-actin networks circumferentially oriented around MD sites and local kinesin/dynein-dependent pulling of newly forming daughter mitochondrial poles) (42). This membrane tension and remodeling when combined with the contractile activity of Drp1 at pre-constricted MD sites is thought to bring the opposing inner monolayers of the constricted membrane tube closer toward hemifission, and ultimately, leak-free fission. However, this view, in our opinion, is rather limited and does not envisage the coordinated interactions of Drp1 with the various cytoskeletal elements and their outcomes in the highly regulated process of MD. How these differential cytoskeletal interactions are staged in a strictly hierarchical manner also warrants a clearer and deeper explanation. Based on emergent data, we propose an alternative, though not mutually exclusive, model of MD that integrates cooperative Drp1-cytoskeleton interactions at MD sites with essential contributions from underlying lipid matrix enriched in curvature-preferring, nonbilayer-prone lipids such as cardiolipin (CL) in the progressive remodeling of the membrane.

The VD is primarily an auto-inhibitory domain that restricts premature Drp1 self-assembly in solution and unregulated (unprimed) association with Mff on membranes (43-45). Based on the available data, we propose that PTM-regulated VD-MT electrostatic interactions at ER-mitochondria contact sites (MD sites) alleviate the auto-inhibition imposed by the VD on Drp1 self-assembly. This MT-modulated ‘conformational priming’ of Drp1 likely facilitates subsequent Drp1 association with F-actin (and associated septins), as F-actin appears to selectively bind assembly-competent higher-order oligomers of Drp1 (34,35). This is likely misconstrued as higher-order Drp1 complexes selectively and directly recruited from the cytosol onto the mitochondrial surface (34,46). MT and F-actin thus may cooperate to generate a ‘readily releasable pool’ (or ‘reservoir’) of VD-rearranged, assembly-primed Drp1 for effective Mff association and local membrane remodeling.

Several lines of evidence indicate that MD site-localized F-actin networks cooperate with accessory actin-binding proteins (ABPs) and septin filaments to act primarily as ‘diffusion barriers’ for both target lipids (e.g., CL, PA) and the MD fission machinery, including Drp1 and its receptor, Mff. This protein-lipid cooperativity, we suggest, induces and sustains the high local membrane curvature required for membrane fission as outlined below. Bundled actin filaments have been shown to run perpendicular to the long axis of the MD site and parallel to the membrane surface (24,47) (Fig. 1C), exhibiting neither end-on abutting nor a circumferential encapsulation of the MD site, dismissing the involvement of protrusive or contractile ring-based forces in membrane constriction (24,47). Instead, ABPs cofilin and cortactin, which localize to MD sites (13) and can independently bind membranes (48), likely anchor and remodel F-actin parallel to the mitochondrial surface (Fig. 1C). We propose that such ABP-anchored F-actin networks sequester negatively charged lipids and elicit membrane phase separation as demonstrated previously with synthetic membranes (49). Controlled F-actin polymerization and dynamics on the membrane surface as shown before (50), coupled with the sequestration of negative curvature-preferring target lipids (CL/PA) by the F-actin/ABP protein lattice, likely initiates the local membrane curvature to support progressive membrane remodeling. At this juncture, F-actin synergizes with Mff to induce curvature-stabilizing Drp1 helical self-assembly to constrict the MD site further (33,34). Concomitant stimulation of Drp1 GTPase activity by a high spatial density of negatively charged, nonbilayer-prone lipids (e.g. fission-favoring CL, which strongly binds the MT-primed VD) augments Drp1 mechanoenzymatic membrane constriction toward hemifission, and ultimately, cooperative cytoskeleton-directed fission (32,33). Septin polymers that interact with F-actin and MT as well as with Drp1 and lipids potentially serve as ‘gatekeepers’ or ‘corrals’ for the above processes and provide a spatial ‘conduit’ that channels MT-primed soluble Drp1 toward cooperative F-actin/ABP/Mff/lipid-directed helical self-assembly at MD sites (Fig. 1C). Thus, we predict that the cytoskeletal elements collectively function as modulators, mediators, and effectors of Drp1-catalyzed MD.

Conclusions

It is becoming increasingly clear that a dynamic cytoskeleton plays an essential role in many, if not all, membrane remodeling processes in cooperation with various mechanoenzymatic DSPs or other membrane-reshaping proteins (e.g., BAR domain-containing proteins). In addition, the target membrane lipid matrix appears to play an active role in these events by promoting and sustaining the requisite high membrane curvature. Yet, the intricate molecular mechanisms underlying such protein-protein and protein-lipid cooperativity, and regulation thereof, remain obscure and await critical experimental dissection. Exciting discoveries surely lie ahead.